Ph.D., University of Arizona

Contact Information
phone: (434) 982-2067
fax: (434) 982-2972
e-mail: rk@virginia.edu

Research Interests
Bayesian theory is the branch of statistics in which my research is rooted and to which it contributes. The broad objective of the theory is to provide principles and procedures for rational decision making under conditions of uncertainty. Thus from a Bayesian perspective, statistical modeling, analysis, estimation and inference are guided by the needs of optimal decision procedures.

My current work centers on what I call Bayesian forecast-decision theory. Its aim is to develop a coherent set of tools for modeling and solving problems at the interface between forecasting and decision making. Among these problems are: elicitation of judgmentaldistributions from experts, modeling of multivariate likelihood functions that characterize the stochastic dependence between forecasts and predictands, updating prior distributions based on forecasts, combining forecasts coming from different sources (experts, models, sensors), statistical verification of forecasts, sufficient comparison of forecasters, optimal decision making based on forecasts (static and sequential decision procedures), and assessment of the economic value of forecasts.

The applied setting that motivates this work and provides data is that of forecasting extreme natural events (such as heavy rains, floods, tornadoes, hurricanes) and issuing warnings of impending hazards to the public. One hopes that statistics can enhance forecasting methods and decision procedures and thereby reduce the societal losses. The theory is general, however, and could be applied to other areas such as investment, industrial, or medical decision making.